Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
126 tokens/sec
GPT-4o
28 tokens/sec
Gemini 2.5 Pro Pro
42 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
38 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

MoVie: Visual Model-Based Policy Adaptation for View Generalization (2307.00972v3)

Published 3 Jul 2023 in cs.LG, cs.CV, and cs.RO

Abstract: Visual Reinforcement Learning (RL) agents trained on limited views face significant challenges in generalizing their learned abilities to unseen views. This inherent difficulty is known as the problem of $\textit{view generalization}$. In this work, we systematically categorize this fundamental problem into four distinct and highly challenging scenarios that closely resemble real-world situations. Subsequently, we propose a straightforward yet effective approach to enable successful adaptation of visual $\textbf{Mo}$del-based policies for $\textbf{Vie}$w generalization ($\textbf{MoVie}$) during test time, without any need for explicit reward signals and any modification during training time. Our method demonstrates substantial advancements across all four scenarios encompassing a total of $\textbf{18}$ tasks sourced from DMControl, xArm, and Adroit, with a relative improvement of $\mathbf{33}$%, $\mathbf{86}$%, and $\mathbf{152}$% respectively. The superior results highlight the immense potential of our approach for real-world robotics applications. Videos are available at https://yangsizhe.github.io/MoVie/ .

Definition Search Book Streamline Icon: https://streamlinehq.com
References (36)
  1. Look where you look! saliency-guided q-networks for generalization in visual reinforcement learning. NeurIPS, 2022.
  2. Unsupervised learning of visual 3d keypoints for control. In ICML, 2021.
  3. Robonet: Large-scale multi-robot learning. CoRL, 2019.
  4. Reinforcement learning with neural radiance fields. arXiv, 2022.
  5. Test-time training with masked autoencoders. NeurIPS, 2022.
  6. Mastering atari with discrete world models. arXiv, 2020.
  7. Self-supervised policy adaptation during deployment. ICLR, 2021.
  8. Modem: Accelerating visual model-based reinforcement learning with demonstrations. ICLR, 2023.
  9. Stabilizing deep q-learning with convnets and vision transformers under data augmentation. NeurIPS, 2021.
  10. Generalization in reinforcement learning by soft data augmentation. In International Conference on Robotics and Automation, 2021.
  11. Temporal difference learning for model predictive control. ICML, 2022.
  12. On pre-training for visuo-motor control: Revisiting a learning-from-scratch baseline. ICML, 2023.
  13. Spatial transformer networks. NeurIPS, 2015.
  14. Reinforcement learning with augmented data. NeurIPS, 2020.
  15. Network randomization: A simple technique for generalization in deep reinforcement learning. arXiv preprint arXiv:1910.05396, 2019.
  16. 3d neural scene representations for visuomotor control. In CoRL, 2022.
  17. Playing atari with deep reinforcement learning. arXiv, 2013.
  18. Human-level control through deep reinforcement learning. Nature, 2015.
  19. Sim-to-real transfer of robotic control with dynamics randomization. In ICRA, 2018.
  20. Learning complex dexterous manipulation with deep reinforcement learning and demonstrations. RSS, 2018.
  21. Bayessim: adaptive domain randomization via probabilistic inference for robotics simulators. arXiv, 2019.
  22. Rrl: Resnet as representation for reinforcement learning. ICML, 2021.
  23. Self-supervised disentangled representation learning for third-person imitation learning. In IROS, 2021.
  24. Third-person visual imitation learning via decoupled hierarchical controller. NeurIPS, 2019.
  25. Third-person imitation learning. ICLR, 2017.
  26. The distracting control suite–a challenging benchmark for reinforcement learning from pixels. arXiv, 2021.
  27. Online learning of unknown dynamics for model-based controllers in legged locomotion. RA-L, 2021.
  28. Test-time training with self-supervision for generalization under distribution shifts. In ICML, 2020.
  29. Deepmind control suite. arXiv, 2018.
  30. Domain randomization for transferring deep neural networks from simulation to the real world. In IROS, 2017.
  31. 3d-oes: Viewpoint-invariant object-factorized environment simulators. arXiv, 2020.
  32. Improving generalization in reinforcement learning with mixture regularization. NeurIPS, 2020.
  33. Learning vision-guided quadrupedal locomotion end-to-end with cross-modal transformers. ICLR, 2021.
  34. Mastering visual continuous control: Improved data-augmented reinforcement learning. arXiv, 2021.
  35. Pre-trained image encoder for generalizable visual reinforcement learning. NeurIPS, 2022.
  36. Visual reinforcement learning with self-supervised 3d representations. RA-L, 2023.
Citations (9)

Summary

We haven't generated a summary for this paper yet.

Github Logo Streamline Icon: https://streamlinehq.com

GitHub